A preliminary seismic study of Taal Volcano, Luzon Island Philippines

S.-H. You a, Y. Gung a,, C.-H. Lin b, K.I. Konstantinou c, T.-M. Chang d, E.T.Y. Chang e, R. Solidum faDepartment of Geosciences, National Taiwan University, Taipei, Taiwanb Institute of Earth Sciences, Academia Sinica, Taipei, TaiwancDepartment of Earth Sciences, National Central University, Taoyuan, TaiwandNational Center for Research on Earthquake Engineering, Taipei, Taiwane Institute of Oceanography, National Taiwan University, Taipei, Taiwanf Philippine Institute of Volcanology and Sesimology, Quezon, Philippines

a r t i c l e i n f o

Article history:Available online 7 November 2012

Keywords:Taal VolcanoSeismicity1-D velocity modelNoise cross-correlation

a b s t r a c t

The very active Taal Volcano lies in the southern part of Luzon Island only 60 km from Manila, the capitalof the Philippines. In March 2008 we deployed a temporary seismic network around Taal that consisted of8 three-component short period seismometers. This network recorded during the period from March toNovember 2008 about 1050 local events. In the early data processing stages, unexpected linear drifting ofclock time was clearly identified for a number of stations. The drifting rates of each problematic stationwere determined and the errors were corrected before further processing. Initial location of each eventwas derived by manually picked P-/S-phases arrival times using HYPO71 and a general velocity modelbased on AK135. Since the velocity structure beneath Taal is essentially unknown, we used travel timesof 338 well-located events in order to derive a minimum 1D velocity model using VELEST. The resultinglocations show that most events occurred at the shallow depth beneath the Taal Volcano, and two majorearthquake groups were noticed, with one lying underneath the western shore of Taal lake and the otherone spread around the eastern flank of the Taal Volcano. Since there is no reported volcano activities dur-ing the operation period of our seismic array, we are still not confident to interpret these findings interms of other natures of volcano at the current stage. However, our work represents an important pio-neer step towards other more advanced seismic studies in Taal Volcano.

! 2012 Elsevier Ltd. All rights reserved.

1. Introduction

Taal Caldera (!20 " 30 km) is located 60 km south of MetroManila on Luzon Island. Inside the caldera is Taal Lake, and TaalVolcano Island is located at the center of the lake. The Taal VolcanoIsland is a !5 km diameter post-caldera structure composed ofseveral eruption centers. Taal Volcano is a 311-m-high stratovol-cano with a lake (Main Crater Lake, 1.2 km wide and 75 m deep)filling the central crater.

Taal Volcano is one of the 15 most dangerous Decade Volca-noes that represent especially high potential hazards to nearbypopulation centers (e.g., Zlotnicki et al., 2009). At least 33 eruptionswere recorded since its first documented activity in 1572, amongwhich the most well-known one is the eruption that occurred in1965. During this VEI 4 event, pyroclastic surges that originatedfrom the southwest flank of the island killed about 200 people(Moore et al., 1966).

The subsurface plumbing activities beneath the volcanic areaare manifested through various geochemical and geothermal

observations, and are usually accompanied by earthquakes. Activ-ities of Taal Volcano have been attentively monitored by the Phil-ippine Institute of Volcanology and Seismology (PHIVOLCS). In theearly 90s, three seismic swarms were reported in Taal area: in early1991, in February 1992 and March 1994. The February 1992 activ-ity was accompanied by a rapid ground deformation at a rate ofabout 1020 cm uplift in a day, and active fissures opened alongthe EW northern flank during the 19921994 seismic activity.Thus, it has been considered to be in a possible state of unrest since1994, as seen from seismic swarms, ground deformation, geyseringactivity and changes in temperature and chemistry (http://www.phivolcs.dost.gov.ph). Studies based on GPS data from19981999 indicate that the largest deformation correlates in timewith anomalous bursts of hydrothermal activity and high-fre-quency local seismicity, suggesting that both deformation and seis-micity are responding to migration of hydrothermal fluids (Bartelet al., 2003). More recently, seismicity in Taal Volcano increasedfrom September 2004. A seismic swarm with a maximum of 20events in a day was recorded on 13 February 2005. After February2005, seismicity slowly decreased until new record of seismicswarms in December 2005. Results from multidisciplinary obser-vations combining electromagnetic, geochemical and thermal

1367-9120/$ - see front matter ! 2012 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.jseaes.2012.10.027

Corresponding author.E-mail address: ycgung@ntu.edu.tw (Y. Gung).

Journal of Asian Earth Sciences 65 (2013) 100106

Contents lists available at SciVerse ScienceDirect

Journal of Asian Earth Sciences

journal homepage: www.elsevier .com/locate / jseaes

survey, suggest that the northern flank located between the craterrim and the 19921994 fissures is connected with a deep thermalsource in Main crater and is reactivated during seismic activity(Zlotnicki et al., 2009).

Various seismological approaches have been used in volcanostudies. Analysis of seismicity and seismic tomography can provideinformation about structure and evolution of magma chambers(e.g. Sherburn et al., 1998; Tanaka et al., 2002; Konstantinouet al., 2007), which may even be used to forecast volcanic eruptions(e.g. Harlow et al., 1996; Bryan and Sherburn, 1999; Jones et al.,2001). Properties of long period (LP) seismic signals, characterizedby single or multiple impulse-like excitations in the frequency do-main can be used to infer the status of volcanoes (e.g. Chouet,1996; McNutt, 1996; Kumagai et al., 2002). Lately, it has beenshown that the Greens function of elastic waves between two seis-mic stations resembles the cross-correlation functions (CCFs) oftheir noisy continuous records (e.g., Weaver and Lobkis 2001,2002; Shapiro and Campillo, 2004; Snieder, 2004). Structure per-turbations induced by volcanic activities can thus have an impacton the noise-derived empirical Greens functions (EGFs). The ideahas been applied to the detection of the subsurface magma activi-ties by examining the robust coda train tailing the major signal ofEGFs (Brenguier et al., 2008).

Given the well documented active seismicity in Taal Volcano, nofurther seismic investigations have been employed in this area,since there was no local seismic network deployed until late March2008.

In this study, with local earthquakes recorded by the newly de-ployed seismic network, we aim to construct the first 1-D model ofseismic velocities for the area around Taal Volcano. A reliable 1-Dmodel is a precondition for most other seismic studies, such as 3-Dtomography or more accurate determination of locations of earth-quake hypocenters. On the other hand, in a parallel project usingthe same data set, we also examine the temporal variations ofnoise-derived EGFs, to explore the potential subsurface perturba-tions induced by volcanic activities, and to ensure the timing qual-ity of the seismic data. Unexpected linear time-drifting of internalclock were noticed from noise-derived daily CCFs. The clock errorshave been further confirmed by teleseismic signals, and were cor-rected before further data processing for the development of the 1-D model.

2. Temporary seismic array

To investigate seismicity and structure beneath the Taal Cal-dera, we deployed a temporary seismic array consisting of sevenstations (TV01TV07) in late March 2008. Four of them were de-ployed on the Taal Volcano Island, the other three on lakeshorearound the Taal Lake. On 21 July 2008, one more station (TV08)was added in northern lakeshore, near the PHIVOLCS observatory.Each station is equipped with a short-period three-componentseismometer (Lennartz Electronic LE-3D Lite 1 Hz) and a GPS syn-chronized clock. The data are continuously recorded and sampled

12054' 12100' 12106'

1354'

1400'

1406'

0 250 500 750 1000Elevation (m)

TV01

TV02

TV03TV04

TV05

TV06

TV07

TV08 5 km

120 121 122

13

14

15Manila

Fig. 1. Station locations (shown here by solid squares) of Taal seismic array. Thesmall box in the upper-left panel shows the study region, in which Manila city isalso shown by a solid triangle.

Table 1Station information of Taal seismic Array.

Station Longitude Latitude Elevation (m) Deployed time

TV01 121.0020 14.0341 19 2008/03/25TV02 121.0117 13.9979 27 2008/03/25TV03 121.0756 13.9621 11 2008/03/25TV04 120.9784 13.9813 40 2008/03/26TV05 120.9729 14.0121 14 2008/03/26TV06 121.0700 14.0435 137 2008/03/26TV07 120.9161 13.9814 321 2008/03/26TV08 120.9867 14.0843 19 2008/07/21

180

210

240

270

300

Julia

n Da

y

-40 -30 -20 -10 0 10 20 30 40

(a) Time lag (sec)

I

II

210

240

270

300

Julia

n Da

y

-40 -30 -20 -10 0 10 20 30 40

(b)

I

II

180

210

240

270

300

Julia

n Da

y

-40 -30 -20 -10 0 10 20 30 40

(c)

I

II

-1.0 -0.5 0.0 0.5 1.0Normalized amplitude

Fig. 2. Examples for clock errors revealed by noise CCFs. Daily CCFs of pairs TV04TV02 (a), TV05TV04 (b), and TV06TV04 (c) are presented here. The waveform ofCCFs is shown in terms of scaled gray pixel with amplitudes normalizedindividually. The symmetric centers of CCF pairs in (a) and (c) clearly drift awayfrom the zero time lag. The star symbols on the right side of each panel mark theoccurring time of teleseismic events discussed in Fig. 3, with I for event 2008/09/08 (Fig. 3b), and II for event 2008/09/29 (Fig. 3c).

S.-H. You et al. / Journal of Asian Earth Sciences 65 (2013) 100106 101

at 100 Hz with a digital recorder (Tokyo Sokushin digital recorderSAMTAC-801H) in all stations. The locations and relevant informa-tion of the seismic array are presented in Fig. 1 and Table 1.

3. Timing errors

Data recorded from late March 2008 to early November 2008are used in this study. Besides the event identification and phasepicking for the development of a seismic model, the continuousdata of the same time period has also been used in a separatestudy, in which we examine the CCFs derived by continuous 1-bit data (e.g., Shapiro and Campillo, 2004) of station pairs, aimingto explore the temporal variations in EGFs which are possibly re-lated to the crust perturbation induced by volcanic activities.

TV01TV02TV03TV04TV06TV07a

2008/05/12, MW:7.0, Distance:2613 km

TV01TV02TV03TV04TV05TV06TV07TV08b

2008/09/08, MW:6.9, Distance:5908km

-10 0 10 20 30 40 50 60TV01TV02TV03TV04TV05TV06TV07TV08c

2008/09/29, MW:6.9, Distance:8171km

Original Record

d

-10 0 10 20 30 40 50 60

e

Time-corrected Record

Fig. 3. The vertical component waveform of three teleseismic events recorded by the Taal array. (a): 2008/05/12, 06:28:01.57, 31.002"N, 103.322"E, depth: 36 km, MW7.0,Sichuan, China. (b): 2008/09/08, 18:52:06.97, 13.401"S, 166.967"N, depth: 110 km, MW 6.9, Vanuatu Islands. (c) 2008/09/29, 15:19:31.59, 29.756"S, 177.683"W, depth:35 km, MW: 6.9, Kermadec Islands, New Zealand. The black line in each panel represents the predicted P arrival time for the central location of array from each event. In bothevents (b) and (c) the accumulated timing errors in TV02 and TV06 are evident. Records shown in panels (d) and (e) are time-corrected results for event (b) and (c)respectively.

Table 2Time-drifting rates determined by linear regression. T is thepredicted drifting time, T0 the time in the data record and DD thenumber of drifting days. The standard deviations (the numberbehind ) for each station are also shown.

Station Time-drifting trend

TV01 T = T0 # 0.015 " DD 0.015TV02 T = T0 # 0.068 " DD 0.063TV03 T = T0 # 0.022 " DD 0.015TV06 T = T0 # 0.140 " DD 0.015TV07 T = T0 # 0.010 " DD 0.018TV08 T = T0 # 0.053 " DD 0.050

Fig. 4. Comparison of the reference CCF (solid line) and the stacking of time-corrected CCFs for station pair TV06TV04. The correlation coefficient (0.993) between two tracesis also shown in the upper-right corner.

102 S.-H. You et al. / Journal of Asian Earth Sciences 65 (2013) 100106

Moreover, the examination of noise-derived EGFs may help toensure the data quality. Basically, two types of instrument errorscan be identified from the noise-derived EGFs: the polarity reversaland time shift of the internal clock. The former is characterized bythe reverse phase in waveform as compared to the reference CCFswith correct polarity and the later is accompanied by a time shift ofthe symmetry between causal (time positive portion of CCF) andacausal (time negative portion of CCF) signals (e.g., You et al.,2010; Lukac et al., 2009).

During the inspection of the noise-derived CCFs, non-negligibletime drift of internal clocks of some stations is noticed. Two exam-ples are shown in Fig. 2, in which daily CCFs are aligned for the en-tire time period, and their amplitudes are normalized to emphasizethe gradual shifting of symmetric center. The seemingly lineartime-drift in station pair TV04TV02 (Fig. 2a) and TV06TV04(Fig. 2c) are unambiguously due to the timing errors of internalclock. The errors could be introduced by one or both stations ofthe problematic station pair. To resolve the issue, we first lookfor reference stations which suffer no such error, as the pairTV04TV05 shown in Fig. 2b, and compare other CCFs paired byone reference station and one suspected station. Results fromFig. 2ac thus suggest station TV02 and TV06 are the ones withclock error among the four stations shown here.

From the above results and station maintenance log, we con-clude that the clock malfunction started on the day after the sta-tion maintenance, i.e., July 2123, 2008. However, the exactcause of time drifting remains unknown.

To further verify the clock errors, we examine earthquake sig-nals from teleseismic events (Fig. 3). The first event occurred onMay 12, 2008 (Fig. 3a), time before the clock malfunction began,and the comparison of P-wave arrival time of each station showsno anomalous feature. The second event occurred on September9, 2008 (Fig. 3b), and the third on September 29, 2008 (Fig. 3c).In both events, the accumulated timing errors in the fore-men-tioned TV02 and TV06 are obvious, and the comparison also dem-onstrate other stations, such as TV03 and TV08, may have similarproblem.

The timing errors are apparently too large to be ignored, andproper correction needs to be done before further data processing.To evaluate the time-drifting of each station, stackings of CCFsfrom the data of the first month are used as reference CCF, forthe fact that no clock error was observed in that period as seenfrom daily CCFs. We then cross-correlate other daily CCFs and their

Table 3The 14-layer initial velocity model.

Depth (km) Vp (km/s) Vs (km/s)

#1.0 3.80 2.202.0 4.30 2.514.0 4.80 2.836.0 5.30 3.148.0 5.80 3.46

10.0 5.91 3.5212.0 6.03 3.5914.0 6.14 3.6516.0 6.26 3.7218.0 6.38 3.7820.0 6.50 3.8525.0 7.01 4.0630.0 7.53 4.2735.0 8.04 4.48

0

5

10

15

20

25

30

35

S-wa

ve tr

avel

tim

e

0 5 10 15 20P-wave travel time

Vp/Vs ratio = 1.72

Fig. 5. The Wadati diagram based on results of HYPO71. A regional Vp/Vs ratio of1.72 is determined from the least square linear fitting for P-wave versus S-wavetravel times.

Fig. 6. 14-Layer 1-D models (black lines) of Vp and Vs determined by VELEST. Thestarting models are shown in gray lines. The result suggests that velocityboundaries at 10 km, 14 km, 18 km and 20 km are not necessary, and there is onlyminor adjustment for depth below 25 km. The mean square residual for eachiteration is shown in the lower-left panel.

0

5

10

15

20

25

30

35

40

45

50

Dep

th (k

m)

0 1 2 3 4 5 6 7 8Velocity (km/s)

: initial P-vel

: initial S-vel: final P-vel: final S-vel

0.00

0.02

0.04

0.06

0.08

0.10M

ean

squa

re re

sidua

ls0 5 10 15

Iterations

Fig. 7. 9-Layer 1-D models (black lines) of Vp and Vs determined by VELEST. Thestarting models are shown in gray lines. The mean square residual for each iterationis shown in the lower-left panel.

S.-H. You et al. / Journal of Asian Earth Sciences 65 (2013) 100106 103

reference CCFs in the frequency range from 0.2 Hz to 0.5 Hz. Theselected frequency band is based on the effective energy distribu-tion of the CCFs amplitude spectrum. The results suggest that allthe time-drifting appear to be linear. We thus inverted for thedrifting rate of each station using least squares linear regression.The drifting rates and the corresponding standard deviations foreach estimate are listed in Table 2. The standard deviations range

from 0.015 to 0.063 s, which are small enough for the purpose ofthis study.

To further assess the accuracy of the time correction scheme,we examine the time-corrected teleseismic signals and CCFs. Asis clearly demonstrated (Fig. 3d and e) that the teleseismic arrivalsare consistent with each other after time corrections. For the time-corrected CCFs, we compare their stacking with the reference CCFin Fig. 4, where both waveforms are almost identical, and a veryhigh correlation (0.993) between them is achieved. Thus, we haveconfirmed that the derived drifting rates are reliable. The clock er-rors of each station are adjusted accordingly prior to the inversion.

4. Minimum 1-D velocity model

Three P arrival pickings is the minimum requirement for thedetermination of hypocenters using HYPO71 (Lee and Lahr,1972). For the time period considered, 1050 recorded events arequalified and their first arrivals of P and S waves are manuallypicked. First, initial locations of these events were determined byHYPO71 with a 14-layer starting model (Table 3). The thicknessof each layer is 2 km in the top 20 km, and 5 km in the depth rangefrom 20 to 35 km, below which velocities of uppermost AK135(Kennett et al., 1995) is used as a half space layer. The model isbased on global continental average of AK135, from which veloci-ties at each layer are linearly interpolated with 3.8 km/s for P and2.2 km/s on the surface layer. Although we aim to build a simpler1-D model, many thin layers are used in the starting model, asthere is no a priori information about how to set up the properboundary depth for velocity profile. Depending on the results ofthe 1-D model inversion, layers with similar velocities will bemerged.

In the first run of HYPO71, the Vp/Vs ration 1.73 is used. Withthe derived initial locations, the Vp/Vs ratio 1.72 is then

Table 4The 9-layer initial velocity model.

Depth (km) Vp (km/s) Vs (km/s)

#1.0 3.80 2.202.0 4.30 2.514.0 4.80 2.836.0 5.30 3.148.0 5.80 3.46

12.0 6.03 3.5916.0 6.26 3.7225.0 7.01 4.0635.0 8.04 4.48

Table 5Final velocity determined by VELEST.

Depth (km) Vp (km/s) Vs (km/s)

#1.0 3.06 2.082.0 4.32 2.564.0 4.72 2.866.0 5.15 3.138.0 5.92 3.42

12.0 6.44 3.4216.0 6.68 3.9725.0 7.14 3.9935.0 8.04 4.47

120 120.4 120.8 121.2 121.6

13.2

13.6

14

14.4

14.8

0

10

20

30

40

Dep

th(km

)

120.0 120.4 120.8 121.2 121.6

W E13.2

13.6

14.0

14.4

14.8

0 10 20 30 40

Depth(km)

S

N

Fig. 8. Map of final hypocenter locations determined from the VELEST.

104 S.-H. You et al. / Journal of Asian Earth Sciences 65 (2013) 100106

determined based on Wadati diagram (Kisslinger and Eengahl,1973) (Fig. 5). Only events with root mean square (RMS) residualsless than 0.25 are used to derive the minimum 1-D velocity model(Kissling, 1988). This criterion let to the rejection of about 2/3 of allevents. The remaining qualified data set consists of 1391 P-waveand 973 S-wave arrivals recorded from 338 earthquakes.

The minimum 1-D velocity model was proposed by Kissling(1988), who also developed the computer package VELEST (Kis-sling, 1995) to solve for the minimum 1-D velocity model froma set of local earthquakes. It represents an ideal 1-D velocity modelobtained by the damped least-square method that takes into ac-count station corrections. For details on the methodology used inVELEST and processing procedure, the reader is referred to Kissling(1995).

The resulting model from the 14-layer starting model is shownin Fig. 6. The result suggests that velocity boundaries at 10 km,14 km, 18 km and 20 km are not necessary. For the larger depths,from the facts that most events are shallow (

generated by fluid transfer processes, such as low/mixed frequencyevents and possibly volcanic tremor. Further analysis of these sig-nals in terms of location and source properties is likely to shedlight on how the magmatic system beneath Taal works and alsohelp towards volcanic hazard assessment. Finally, as more dataaccumulate it will be possible to use local earthquake tomographyin order to investigate the 3D velocity structure of the volcano andpinpoint any locations of partial melt.

Acknowledgements

We would like to thank the Philippine Institute of Volcanologyand Seismology (PHIVOLCS) for their assistance in the deploymentand maintenance of the seismic network. This research is sup-ported by the National Science Council of Taiwan under the grantsNSC 100-2116-M-002-025 and NSC 100-2119-M-001-020.

References

Bartel, B.A., Hambuger, M.W., Meertens, C.M., Lowry, A.R., Corpuz, E., 2003.Dynamics of active magmatic and hydrothermal systems at Taal Volcano,Philippines, from continuous GPS measurements. Journal of GeophysicalResearch 108 (B10), 2475. http://dx.doi.org/10.1029/2002JB002194.

Bryan, C.J., Sherburn, S., 1999. Seismicity associated with the 19951996 eruptionsof Ruapehu Volcano, New Zealand: narrative and insights into physical process.Journal of Volcanology and Geothermal Research 90, 118. http://dx.doi.org/10.1016/S0377-0273(99)00016-5.

Brenguier, F., Shapiro, N.M., Campillo, M., Ferrazzini, V., Duputel, Z., Coutant, O.,Nercessian, A., 2008. Towards forecasting volcanic eruptions using seismicnoise. Nature 1, 126130. http://dx.doi.org/10.1038/ngeo104.

Chouet, B.A., 1996. Long-period volcano seismicity: its source and use in eruptionforecasting. Nature 380, 309316.

Harlow, D.H., Laguerta, J.A., Power, Laguerta, E.P., Ambubuyog, G., White, R.A.,Hoblitt, R.P., 1996. Precursory seismicity and forecasting of the June 15, 1991,eruption of Mount Pinatubo. In: Punongbayan, R.S., Newhall, C.G. (Eds.), Fireand Mud Eruptions and Lahars of Mount Pinatubo, Philippine. University ofWashington Press, Seattle.

Jones, J.P., Thurber, C.H., Lutter, W.J., 2001. High precision location of pre-eruptionseismicity at Mount Pinatubo, Philippines, 30 May3 June, 1991. Physics of theEarth and Planetary Interiors 123, 221232. http://dx.doi.org/10.1016/s0031-0201(00)00211-9.

Kennett, B.L.N., Engdahl, E.R., Buland, R., 1995. Constraints on seismic velocities inthe Earth from travel times. Geophysical Journal International 122, 108124.

Kisslinger, Carl., Eengahl, E.R., 1973. The interpretation of the Wadati diagram withrelaxed assumptions. Bulletin of the Seismological Society of America 63, 17231736.

Kissling, E., 1988. Geotomography with local earthquake data. Reviews ofGeophysics 26, 659698.

Kissling, E., 1995. Velest Users Guide, International report Institute of Geophysics,ETH Zurich.

Kumagai, H., Chouet, B.A., Nakano, M., 2002. Temporal evolution of a hydrothermalsystem in Kusatsu-Shirane Volcano, Japan, inferred from the complexfrequencies of long-period events. Journal of Geophysical Research 107 (B10),2236. http://dx.doi.org/10.1029/2001JB000653.

Konstantinou, K.I., Lin, C.-H., Liang, W.-T., 2007. Seismicity characteristics of apotentially active quaternary volcano: the Tatun Volcano Group, northernTaiwan. Journal of Volcanology and Geothermal Research 160, 300318. http://dx.doi.org/10.1016/j.jvolgeores. 2006.09.009.

Lee, W.H.K., Lahr, J.C., 1972. Hypo71: a computer program for determininghypocenter, magnitude, and fist motion pattern for local earthquakes. OpenFile Report 75311, US Geological Survey.

Lukac, M., Davis, P., Claytion, R., Estrin, D., 2009. Recovering temporal integrity withdata driven time synchronization. In: Proceedings of the 2009 InternationalConference on Information Processing in Sensor Networks 00. pp. 6172.

McNutt, S.R., 1996. Seismic monitoring and eruption forecasting of volcanoes: areview of the state-of-the-art and case histories. In: Scarpa, Tilling (Ed.),Monitoring and Mitigation of Volcanic Hazards. Springer-Verlag, Berlin.

Moore, J.G., Nakamura, K., Alcaraz, A., 1966. The 1965 eruption of Taal Volcano.Science 151, 955960.

Shapiro, N.M., Campillo, M., 2004. Emergence of broadband Rayleigh waves fromcorrelations of the ambient seismic noise. Geophysical Research Letters 31,L07614. http://dx.doi.org/10.1029/2004GL019491.

Sherburn, S., Scott, B.J., Nishi, Y., Sugihara, M., 1998. Seismicity at White IslandVolcano, New Zealand: a revised classification and inferences about sourcemechanism. Journal of Volcanology and Geothermal Research 83, 287312.

Snieder, R., 2004. Extracting the Greens function from the correlation of codawaves: a derivation based on stationary phase. Physical Review E 69, 046610.http://dx.doi.org/10.1103/PhysRevE.69.046610.

Tanaka, S., Hamaguchi, H., Ueki, S., Sato, M., Nakamichi, H., 2002. Migration ofseismic activity during the 1998 volcanic unrest at Iwate Volcano, NortheasternJapan, with reference to P and S wave velocity anomaly and crustaldeformation. Journal of Volcanology and Geothermal Research 113, 399414.http://dx.doi.org/10.1016/S0377-0273(01)00273-6.

Weaver, R.L., Lobkis, O.I., 2001. Ultrasonics without a source. Thermal fluctuationcorrelations at MHz frequencies. Physical Review Letters 87, 134301. http://dx.doi.org/10.1103/PhysRevLett.87.134301.

You, S.-H., Gung, Y., Chiao, L.-Y., Chen, Y.-N., Lin, C.-H., Liang, W.-T., Chen, Y.-L., 2010.Multi-scale ambient noise tomography of short period rayleigh waves acrossNorthern Taiwan. Bulletin of the Seismological Society of America 100, 31653173. http://dx.doi.org.10.1785/0120090394.

Zlotnicki, J., Sasai, Y., Toutain, J.P., Villacorte, E.U., Bernard, A., Julio, P.S., Gordon Jr.,J.M., Corpuz, E.G., Harada, M., Punongbayan, J.T., Hase, H., Nagao, T., 2009.Combined electromagnetic geochemical and thermal surveys of Taal Volcano(Philippines) during the period 20052006. Bulletin of Volcanology 71, 2947.http://dx.doi.org/10.1007/s00445-008-0205-2.

106 S.-H. You et al. / Journal of Asian Earth Sciences 65 (2013) 100106

A preliminary seismic study of Taal Volcano, Luzon Island Philippines1 Introduction2 Temporary seismic array3 Timing errors4 Minimum 1-D velocity model5 Spatial distribution of located earthquakes6 ConclusionsAcknowledgementsReferences